class: center, middle, inverse, title-slide # Visualizing Activity Data from Wearable Devices in R to Create Personalized Activity Reports ## Duke Industry Statistics Symposium ### Stefan Avey <<a href="mailto:savey@its.jnj.com" class="email">savey@its.jnj.com</a>> ### 23 Apr 2021 --- # Fatigue and Sleep Disturbance are Prevalent Symptoms - Digital health technologies, such as activity trackers, may provide a novel way to measure impacts of fatigue and sleep disturbances in patients with chronic immune diseases. <img src="include/fatigue_sleep_ideafast.png"> ??? - Prevalence in many chronic immune-mediated inflammatory diseases is above 50% for fatigue and sleep disturbance - New tools are needed to understand the real impact of these symptoms on patient's daily lives. - This schematic is taken from one of those initiatives, the IMI-funded project IDEA-FAST (Identifying Digital Endpoints of Fatigue and Sleep Disturbance) --- # Introduction to Actigraphy .pull-left[ - Actigraphy is a **non-invasive** method of monitoring human rest/activity patterns through monitoring of **movement**. - Often, a small wrist-watch with a built-in **accelerometer** is used to collect **actigraphy** data. - Example derived data include: step count, time in intense physical activity, sleep time. <img src="include/actigraph_cpiw.jpeg" height="250"> <img src="include/actiwatch2.png" height="200"> ] .pull-right[ <img src="include/cpiw_activity_count_traces.png"> ] ??? - Actigraphy watches have accelerometers which measure movement in 3 axes (x, y, z) - What's shown on the right is a summarized "activity count" which summarizes the amount of movement detected by the accelerometer in 1 minute **epochs** over a full 24-hour day. --- # Why Create Personalized Activity Reports? - We want to **provide value** to study participants with the data we are already collecting - Study participants expect it from consumer wearables. - Typically the data collected in clinical trials is not shared with study participants (prevent unblinding, prevent biasing patient-reported data, QC'ed data not available in real-time) <center> <img src="include/consumer_dashboards.png" height="325"> </center> ??? - We wanted to provide value while also making sure we... - Prevent biasing data - Ensure data privacy - Increase patient engagement (drive better compliance) --- # Activity Report in Phase 0 Study .left-column[ - Ongoing Phase 0 Observational Study including a research-grade actigraphy watch - Built a report and process for data sharing **during the study** - Static PDF report with activity/sleep summary metrics - Providing a report per participant per month ] .right-column[ <img src="include/activity_report_tiles.png" height="450"> ] ??? - Screenshot here shows the first page of the mock report which was delivered as a monthly summary. - Initially had grand plans of making this like consumer-facing app but due to limitations with technology we were using we could not integrate directly into the study mobile app or use HTML / interactive plots. - Had to use static PDF summaries available via the study mobile app or online dashboard as downloads --- # Activity Report in Phase 0 Study .left-column[ - Report is built using Rmarkdown and rendering to PDF - All visualizations are created using R code ] .right-column[ <img src="include/code_tiles.png"> .center[For more details, see [my answer on Stack Overflow](https://stackoverflow.com/questions/47105282/valuebox-like-function-for-static-reports/56841865#56841865)] ] ??? - Create the tiles with `geom_tile()` using the `color` column - Add the numeric values as text in `value` column - Add the labels for each box stored in the `info` column - Use `geom_text()` to add the icons by specifying the unicode symbol. --- # Activity Report in Phase 0 Study .left-column[ - R code to create the daily bar chart of step count ] .right-column[ <img src="include/activity_report_steps.png" height="450"> ] ??? - Snapshot of average of the daily values for the entire month in calendar-like layout - Given only after study participants complete the daily and weekly patient reported outcomes. --- # Activity Report in Phase 0 Study .left-column[ - Static PDF report with activity/sleep summary metrics - Providing 1 report to each study participant per month ] .right-column[ <img src="include/code_step_plot.png"> ] ??? - The code for this plot is very simple in ggplot2 consisting of mainly `geom_bar()` and `geom_text` calls. --- # Automated Creation of Activity Reports with Shiny - Internal R Shiny app was built to facilitate creating these reports <center> <img src="include/shiny_dashboard.png" height = "400"> </center> ??? - Semi-Automated Workflow - Export Data from ActiGraph (actigraphy vendor's) dashboard - Upload data file into Shiny app - Select months and participants - Download a batch of activity reports - Transferred from Janssen to the virtual site to add to participant app/web portal. - Workflow could be automated by pulling data from the vendor's API. --- # Future Directions - Towards Interactivity - Actigraphy data is rich in density and can be displayed in many different ways - More research is needed to understand the best way to visualize the data and what data study participants want to see. - Interactive graphics could be a way to engage participants who really want to dig into the data. --- # Example - Doughnut Charts ```r act_df %>% plot_ly(labels = ~Type, values = ~fraction) %>% add_pie(hole = 0.5) %>% layout(showlegend = TRUE, xaxis = list(showgrid = FALSE, zeroline = FALSE, showticklabels = FALSE), yaxis = list(showgrid = FALSE, zeroline = FALSE, showticklabels = FALSE)) ``` <div id="htmlwidget-8f836166d559454ecd73" style="width:432px;height:288px;" class="plotly html-widget"></div> <script type="application/json" data-for="htmlwidget-8f836166d559454ecd73">{"x":{"visdat":{"17227e964fa5":["function () ","plotlyVisDat"]},"cur_data":"17227e964fa5","attrs":{"17227e964fa5":{"labels":{},"values":{},"alpha_stroke":1,"sizes":[10,100],"spans":[1,20],"type":"pie","hole":0.5,"inherit":true}},"layout":{"margin":{"b":40,"l":60,"t":25,"r":10},"showlegend":true,"xaxis":{"showgrid":false,"zeroline":false,"showticklabels":false},"yaxis":{"showgrid":false,"zeroline":false,"showticklabels":false},"hovermode":"closest"},"source":"A","config":{"showSendToCloud":false},"data":[{"labels":["Asleep","Awake_Immobile","Awake_Active","Awake_HighlyActive"],"values":[0.7,0.05,0.15,0.1],"type":"pie","hole":0.5,"marker":{"color":"rgba(31,119,180,1)","line":{"color":"rgba(255,255,255,1)"}},"frame":null}],"highlight":{"on":"plotly_click","persistent":false,"dynamic":false,"selectize":false,"opacityDim":0.2,"selected":{"opacity":1},"debounce":0},"shinyEvents":["plotly_hover","plotly_click","plotly_selected","plotly_relayout","plotly_brushed","plotly_brushing","plotly_clickannotation","plotly_doubleclick","plotly_deselect","plotly_afterplot","plotly_sunburstclick"],"base_url":"https://plot.ly"},"evals":[],"jsHooks":[]}</script> ??? - Way to show percent of time spent in different (mobility) states - Click around to show they are interactive and not static --- # <!-- NOTE: Update the dashboard to turn off the legend in doughnut plot --> <!-- Can actually include the dashboard itself but doesn't render well --> <iframe src="include/patient_dashboard.html" height="600" width = "1000"></iframe> <!-- <\!-- Show a screenshot of the dashboard instead -\-> --> <!-- <center> --> <!-- <img src="include/dashboard_screenshot.png" height="500"> --> <!-- </center> --> ??? - Can host this on patient community portal - Or can send as a standalone HTML file that allows patients to explore the data - We are currently working to understand what our study participants would want to see --- # Summary - Inclusion of digital health tools, such as actigraphy, are increasingly used in clinical trials to better understand patient's activities of daily living. - In addition to using the data for research, that data can be summarized in a way that is value for study participants. - Showed an example of how we have **visualized actigraphy data** to create **personalized activity reports** in a phase 0 study. - When possible, actigraphy summary data can be shared back with study participants during (or after) the study.